Model reusability in decision support systems : a multi-meta view approach

Abstract

Decision Support Systems (DSS) deal with a variety of models which are developed in an organization by heterogeneous resources: e.g. solver, user interface, and modeling syntax. Providing model reusability is important when decomposition, composition, and parameter change of models are frequently occurred. Model reuse was intended to increase the productivity of DSS modeling. It designates an attempt to reuse the whole or parts of models that already exist in an organization. Model reusability is crucial for the productivity of DSS modeling for the following reasons. First, it is inefficient to formulate models from scratch. Second, different modelers develop almost identical models for similar purposes. This may not lead to wrong decisions, but it does contribute to organizational inefficiency. Third, it is well-established that in order to achieve productivity gains, maintenance must be performed at a higher abstraction level than the code level. If a large amount of target models is specified in code level, then reusing the models would be much more difficult than those of higher abstraction level. Moreover, providing traceability between different abstraction level of models should be supported. Thus, acquiring reusable model constructs from the models that are already scattered in an organization is considered to resolve the above problems. This dissertation proposes an idea of reverse modeling to enhance the reusability of DSS models. Reverse modeling (RM) is a process of identifying and collecting model constructs and translating them into a reusable form. The central idea of RM is based on a meta system that specifies meta model and generates the model translator in order to interpret the constructs embedded in a model. A framework of RM consists of: (1) meta system, (2) model translator, (3) model base, (4) source models, and (5) modelers. The meta level model definition is delineated, which would generate DSS models and model translator. The model tra...